I saw when browsing we can use data augmentation for creating a dataset for face recognition. The augmented images may include inverted, tilted or distorted faces. Do the model detect the face from the inverted image. When I tried my model cant able to detect any inverted or tilted faces.
2
-
$\begingroup$ Did you check your inverted image ? Look at them, be sure your data augmentation function is not producing garbage. $\endgroup$– AstariulOct 25, 2018 at 23:24
-
$\begingroup$ I used augmentor lib for augmentation $\endgroup$– SRJ577Oct 26, 2018 at 6:32
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
This is the purpose of rotational augmentation. In fact you should augment your images at different angle as well than just inverting it. This will make your model angle agnostic.
-
$\begingroup$ I have done than by rotating image in smaller angles. Now model can detect the face. But is there something available to detect faces in the inverted image. $\endgroup$– SRJ577Oct 26, 2018 at 9:50